摘要
提出了一种基于图像分类的图像检索算法。算法首先以图像间的相关系数作为距离进行聚类,并确定每类图像的中心图像。图像检索过程分为两步:首先寻找匹配程度最高的中心图像,然后在该中心图像所在类中寻找最佳匹配图像。由于分类和确定中心均可离线操作,所以该算法显著加快了图像检索速度。此外,本文还运用信息熵分析了分类的有效性。实验结果表明:算法有效。
An efficient method which based on the image classification is being used to the image retrieval . The cross correlation among the images is being used as the distance to cluster ,and the central image of per classification is found . The image retrieval method consists of two phases ： （1） finding the central image which has the best matching in all the central images （2） finding the perfect matching image from the classification which has the central image . The image classification and the finding for the central image are the off-line works , so they accelerate the image retrieval . And the information entropy is also being used to prove the validity of the image classification . The experiment indicates that the method has advantage over other methods .
出处
《微计算机信息》
2009年第15期294-296,共3页
Control & Automation
关键词
图像分类
图像检索
聚类分析
相关系数
image classification
image retrieval
clustering
correlation
